Link between electrode channels and units

[GaelleChapuis]

Hello,

I am using this dataset for testing

dandiset_id = '000409'
filepath = 'sub-KS042/sub-KS042_ses-07dc4b76-5b93-4a03-82a0-b3d9cc73f412_behavior+ecephys+image.nwb'

And loading the electrode channels data like so:

channels_all = nwb.ec_electrodes.to_dataframe()

This session has recordings from 2 Neuropixels probes probe00 and probe01 (each have 384 channels), in AP / LF band.
The channels dataframe contains thus 2 (AP / LF) x 2 (probe00/probe01) x 384 (channels).

I am loading the units data like so, and am filtering to keep those just for a single electrode:

units = nwb.units.to_dataframe()
probe_label = 'probe00'
units_one_probe = units[units.probe_name == probe_label]

Note here that the probe label is probe00.

I am now trying to retrieve the corresponding information in the channels table, i.e. I want to load the information for the AP channels, just for the corresponding probe_label.
I am thus looking for a column in the channels table that would contain the label probe00 or probe01.

The column channels.group_name contains the unique values 'NeuropixelsShank00', 'NeuropixelsShank01'. Can I safely assume that probe00 is 'NeuropixelsShank00' (it would be easier programmatically if the convention was to keep the same key name)? Or is there another place to retrieve this information from (either in channels, or in units)?

Thank you !

[CodyCBakerPhD]

I'd recommend checking out this NWB tutorial on ecephys metadata to get a complete explanation of cross-linkage between the various data types

The short answer to the question is: the electrodes table has a column groups that links each electrode to a single ElectrodeGroup (used to model qualitative structural difference across a probe) which itself must link to a Device (which is lowest level representation of an individual probe in this case - that's also where the trajectory estimate and serial number will go in the next round of conversions).

If you haven't yet, I'd recommend checking out some of these NWB files using the NWB Widgets (pip install nwbwidgets), which visually lays out much of this file structure directly (and has a few neat plots such as PSTH from the untis tab)

from nwbwidgets import nwb2widget

nwb2widget(nwbfile)  # using the same nwbfile you're currently accessing data from